Adding float32 series to dataframe and then attempting to join()

[bburan]

Tested with Pandas 0.7.3

I have a IPython notebook that explores this issue (basically it only seems to occur when you add a float32 series to the DataFrame after creating it, no other datatype I tested seems to trigger this bug). Please contact me if you want the notebook since I don't seem to be able to attach files to this issue. The simplest way to replicate it is via the following lines:

import numpy as np
import pandas
a = np.random.randint(0, 5, 100)
df = pandas.DataFrame({'a': a})
s = pandas.Series(np.random.random(5), name='md')
df.join(s, on='a') # this is OK
df['b'] = np.random.randint(0, 5, 100)
df.join(s, on='a') # this is still OK
df['c'] = np.random.randint(0, 5, 100).astype('f')
df.join(s, on='a') # this fails

The traceback is:

ValueError                                Traceback (most recent call last)
C:\Users\brad\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\Enthought\<ipython-input-1-454edb5f6546> in <module>()
      8 df.join(s, on='a') # this is still OK
      9 df['c'] = np.random.randint(0, 5, 100).astype('f')
---> 10 df.join(s, on='a') # this fails
     11 

C:\Python27\lib\site-packages\pandas\core\frame.pyc in join(self, other, on, how, lsuffix, rsuffix, sort)
   3285         # For SparseDataFrame's benefit

   3286         return self._join_compat(other, on=on, how=how, lsuffix=lsuffix,
-> 3287                                  rsuffix=rsuffix, sort=sort)
   3288 
   3289     def _join_compat(self, other, on=None, how='left', lsuffix='', rsuffix='',

C:\Python27\lib\site-packages\pandas\core\frame.pyc in _join_compat(self, other, on, how, lsuffix, rsuffix, sort)
   3298             return merge(self, other, left_on=on, how=how,
   3299                          left_index=on is None, right_index=True,
-> 3300                          suffixes=(lsuffix, rsuffix), sort=sort)
   3301         else:
   3302             if on is not None:

C:\Python27\lib\site-packages\pandas\tools\merge.pyc in merge(left, right, how, on, left_on, right_on, left_index, right_index, sort, suffixes, copy)
     29                          right_index=right_index, sort=sort, suffixes=suffixes,
     30                          copy=copy)
---> 31     return op.get_result()
     32 if __debug__: merge.__doc__ = _merge_doc % '\nleft : DataFrame'
     33 

C:\Python27\lib\site-packages\pandas\tools\merge.pyc in get_result(self)
     80                                       copy=self.copy)
     81 
---> 82         result_data = join_op.get_result()
     83         result = DataFrame(result_data)
     84 

C:\Python27\lib\site-packages\pandas\tools\merge.pyc in get_result(self)
    495         for klass in kinds:
    496             klass_blocks = [mapping.get(klass) for mapping in blockmaps]
--> 497             res_blk = self._get_merged_block(klass_blocks)
    498             result_blocks.append(res_blk)
    499 

C:\Python27\lib\site-packages\pandas\tools\merge.pyc in _get_merged_block(self, blocks)
    509 
    510         if len(to_merge) > 1:
--> 511             return self._merge_blocks(to_merge)
    512         else:
    513             unit, block = to_merge[0]

C:\Python27\lib\site-packages\pandas\tools\merge.pyc in _merge_blocks(self, merge_chunks)
    545                 com.take_fast(blk.values, unit.indexer,
    546                               None, False,
--> 547                               axis=self.axis, out=out_chunk)
    548 
    549             sofar += len(blk)

C:\Python27\lib\site-packages\pandas\core\common.pyc in take_fast(arr, indexer, mask, needs_masking, axis, out, fill_value)
    264         return take_2d(arr, indexer, out=out, mask=mask,
    265                        needs_masking=needs_masking,
--> 266                        axis=axis, fill_value=fill_value)
    267 
    268     result = arr.take(indexer, axis=axis, out=out)

C:\Python27\lib\site-packages\pandas\core\common.pyc in take_2d(arr, indexer, out, mask, needs_masking, axis, fill_value)
    236             out = np.empty(out_shape, dtype=arr.dtype)
    237         take_f = _get_take2d_function(dtype_str, axis=axis)
--> 238         take_f(arr, indexer, out=out, fill_value=fill_value)
    239         return out
    240     else:

C:\Python27\lib\site-packages\pandas\_tseries.pyd in pandas._tseries.take_2d_axis1_float64 (pandas\src\tseries.c:49365)()

ValueError: Buffer dtype mismatch, expected 'float64_t' but got 'float'

[bburan]

Update: I have tested this with the latest code (0.8.0.dev-d840eda) and the problem is still present.

[wesm]

This has been fixed in the development version, look for 0.9 release out shortly